Ocean and Coastal Management 187 (2020) 105078

Contents lists available at ScienceDirect

Ocean and Coastal Management

journal homepage: http://www.elsevier.com/locate/ocecoaman

Local land subsidence in Miami Beach (FL) and Norfolk (VA) and its
contribution to flooding hazard in coastal communities along the U.S.
Atlantic coast
Simone Fiaschi a, b, *, Shimon Wdowinski c
a
Department of Geosciences, University of Padua, Padua, Italy
b
Currently at UCD School of Earth Science, University College Dublin, Dublin, Ireland
c
Department of Earth and Environment, Florida International University, Miami, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Increasing rate of sea level rise (SLR) along the US Atlantic coast has resulted in increasing flooding hazard in
InSAR several coastal communities, including Boston (MA), Norfolk (VA), and Miami Beach (FL). Here, we evaluate the
Land subsidence contribution of local land subsidence to coastal flooding hazard in two communities, Miami Beach and Norfolk,
Flood hazard
using Interferometric Synthetic Aperture Radar (InSAR) time series observations. The InSAR analysis relies on
Atlantic coastal communities
two data sets of ERS-1/2 scenes that were acquired during 1992–1999. The long period covered by the data sets
and the large number of available scenes (>20), allowed us to detect movements with lower uncertainty levels
(up to 2.4 mm/yr) compared to previous studies. Our results revealed the occurrence of localized subsidence in
both communities. In Miami Beach, subsidence at rates of 1–3 mm/yr occurred in a small portion of the territory,
mainly in parts of the city built on reclaimed wetlands. In Norfolk, relative subsidence occurred in several
localized areas, some along the shore and some inland, at rates of 1–3 mm/yr, while only few sectors show
subsidence up to 6 mm/yr. In these areas, the subsidence is higher and reaches ~8 mm/yr if the combined effects
of regional-scale (~1.7 mm/yr) and InSAR-derived subsidence is considered. The subsidence observed in this
study indicates localized areas of relative higher rate of SLR and a potential increased coastal flooding hazard for
both communities.

1. Introduction
The US Atlantic coast is one of the most vulnerable areas to sea level
rise (SLR) due to its low elevation, large population concentrations, and
economic importance. Many communities along this 3000-km long
coastline have been affected by coastal flooding induced mainly by
storm surge, high tide, and/or heavy rain. Rising sea level increases the
likelihood of such flooding, including rain-induced flooding due to
reduced efficiency of gravity-based drainage systems. Over the past
decade, several coastal communities, including Boston (MA), Norfolk
(VA), and Miami Beach (FL), have experienced increased tidal
(nuisance) flooding frequency, which caused significant property dam­
age and disruptions to daily life (Ezer, 2013; Kirshen et al., 2008;
Kleinosky et al., 2007; Wdowinski et al., 2016). The long-term rates
(>60 years) of relative SLR in Norfolk is 4.6 � 0.5 mm/yr and in Miami
Beach is 2.4 � 0.4 mm/yr, whereas the decadal-scale rates are faster and
are of 5.2 � 0.6 mm/yr in Norfolk and of 9.0 � 4.0 mm/yr in Miami
Beach (Zervas, 2009; Ezer, 2013; Wdowinski et al., 2016). Rising sea
level can also increase the potential damage from storm surge, as in the
case of the 2012 “super-storm” Sandy, which took the lives of 185 people
and caused property and economic damage of about $65 billion (Aerts
et al., 2014).
Vertical land movements (subsidence) can also increase coastal
flooding hazard, as it increases the relative rate of SLR. Land subsidence
occurs at different scales and can reach rates of more than 200 mm/yr, as
observed in Jakarta, Indonesia (Chaussard et al., 2013). The main causes
of land subsidence in the tectonically quiescent US Atlantic coast (pas­
sive margin) include viscoelastic mantle response to melting glaciers in
the Late Pleistocene, ground water extraction, and local sediment
compaction. The regional-scale subsidence due to mantle flow, termed
Glacial Isostatic Adjustment (GIA), affects the Atlantic coast from New
Hampshire to northern Florida (Karegar et al. 2016, 2017; Sella et al.,

* Corresponding author. Department of Geosciences, University of Padua, Padua, Italy.

E-mail address: simone.fiaschi@ucd.ie (S. Fiaschi).

https://doi.org/10.1016/j.ocecoaman.2019.105078
Received 26 September 2019; Received in revised form 8 December 2019; Accepted 17 December 2019
Available online 26 December 2019
0964-5691/© 2019 Elsevier Ltd. All rights reserved.
S. Fiaschi and S. Wdowinski
2007). Basin-scale subsidence has occurred in various locations along
the mid-Atlantic coastal plain, including the southern Chesapeake Bay
region and along the Carolina coasts due to massive groundwater
extraction (Eggleston and Pope, 2013). Local-scale land subsidence oc­
curs in many locations along coastal areas, especially in sediment rich
areas, such as river deltas, and reclaimed land and wetlands. For
example, land subsidence in New Orleans occurred mainly in new
neighbourhoods built on reclaimed wetlands (Dixon et al., 2006).
Coastal subsidence along the US Atlantic coast has been estimated by
both geological and geodetic methods and found to be in the range of
0–4 mm/yr. Geologically-based subsidence rates were determined from
late Holocene (past 4000 years) sediment records in various coastal lo­
cations and found to be in the range of 0–2 mm/yr (Engelhart and
Horton, 2012; Karegar et al. 2016, 2017; Kemp et al., 2014; Nikitina
et al., 2015). The geological rates reflect the long-term subsidence
components of sediments compaction and GIA. Geodetic observations
acquired by levelling, extensometers, and Global Positioning System
(GPS) provide direct measurements of the present-day subsidence rate.
Throughout most of the Atlantic coast, the geodetic and geologic
determined rates are similar, except along a section of the mid-Atlantic
coast between Virginia and South Carolina (32.5–38� N), where
present-day geodetic rates reach 4 mm/yr, about double the long-term
geologic rate (Karegar et al., 2016). The high subsidence rate in the
mid-Atlantic region is attributed to aquifer system deformation in
response to massive groundwater extraction, which began in the turn of
the 20th century, but has intensified since the 1960’s (Eggleston and
Pope, 2013; Karegar et al., 2016) (Fig. S1). The geodetically observed
subsidence rates along the US Atlantic coast are based on GPS and/or
2
Ocean and Coastal Management 187 (2020) 105078
extensometers point measurements located tens of kilometres from one
another. Thus, these measurements provide the long wavelength (>50
km) characterization of coastal subsidence, but are limited in detecting
shorter wavelength subsidence, for example occurring by sediment
compaction. A recent Interferometric Synthetic Aperture Radar (InSAR)
analysis of ALOS-1 data acquired during 2007–2011 over the Hampton
Roads, which also includes the city of Norfolk, VA, detected localized
deformation patterns (<100 m) in that area, but revealed limited results
due to the short length of the time series (Bekaert et al., 2017).
In this study, we use InSAR time series acquired by the ERS-1/2
satellites during 1992–1999, to detect localized subsidence in two
coastal communities, Miami Beach and Norfolk (Fig. 1A). Both com­
munities have been subjected to repeated coastal flooding over the past
decade (Ezer, 2013; Wdowinski et al., 2016). In Miami Beach flooding
has occurred mostly in the western part of the city, west of Alton Road
(Fig. 1B), which was constructed on reclaimed wetland (Wdowinski
et al., 2016). In Norfolk, flooding has occurred in multiple locations
throughout the city characterized by low elevation (Fig. 1C). Our study
relies on 6-7 year-long ERS-1/2 observations, which are significantly
longer than the 4-year long ALOS-1 observations used previously to
study land subsidence in the Norfolk area (Bekaert et al., 2017). The
longer observation period and the larger number of available SAR scenes
enable us to detect small vertical movements with almost half the un­
certainty level compared to the shorter time series of the previous study
(Havazli and Wdowinski, 2017). The high spatial resolution of the InSAR
observations (~20 m by 20 m) is capable of detecting the distribution
and rate of localized subsidence in both study areas and, consequently,
providing better local-scale assessments of coastal flooding hazards for
Fig. 1. Location of the study areas. (A) Map of the
US Atlantic coast showing the locations of Miami
Beach and Norfolk (VA); (B) Flood affected areas in
Miami Beach. Blue dots mark the locations of the
claimed rain and high tide flooding events during
1998–2012. Red lines mark the location of roads
affected by flooding during the same time period
(data from Wdowinski et al., 2016). (C) Norfolk
(VA), the blue lines are the streets and intersections
reported as flooded in storm from 9/30/2010 to
05/01/2015 (data from http://va-norfolk.civicplus.
com). The red lines mark the location of the main
roads discussed in the text. Base image source: Esri,
DigitalGlobe, GeoEye, Earth Geographics, CNE­
S/Airbus DS, USDA, USGS, AeroGRID, IGN, and the
GIS User Community. (For interpretation of the ref­
erences to colour in this figure legend, the reader is
referred to the Web version of this article.)
S. Fiaschi and S. Wdowinski
both communities.
2. Available data sets
Results were obtained from two sets of C-band (5.6 cm wavelength)
SAR scenes acquired by the ERS-1/2 satellites, which are available from
the European Space Agency (ESA) archive. The first data set, covering
the Miami Beach area, consists of 24 scenes acquired from 03/05/1993
to 06/05/1999 (Table S1); the second, covering the city of Norfolk,
consists of 32 scenes acquired from 19/06/1992 to 09/11/1998
(Table S2). Both sets of images have been acquired with average look
angle of 23� , heading angle of 193� and revisit time of 35 days. It must be
noted that, even if the minimum revisit time of the ERS-1/2 satellites is
35 days, the temporal distribution of the SAR scenes is not uniform and
gaps up to two years are present in both data sets (Tables S1 and S2). The
nominal pixel spacing is 7.90 m in slant range and 3.98 m in azimuth.
The ERS-1/2 data extend over a relatively long-time span (6–7 years)
and contain sufficient number of scenes (>20) needed to reach mea­
surement accuracy of 1–2 mm/yr (Havazli and Wdowinski, 2017). Other
satellites missions (ALOS-1/2, RADARSAT-1/2, ENVISAT) did not ac­
quire a sufficient number of scenes along the US Atlantic coast or ac­
quired data over a shorter time period and, hence, cannot provide the
accuracy required for our study.
3. Methods
The two ERS-1/2 data sets were processed using the Small Baseline
Subset (SBAS) algorithm (Berardino et al., 2002) as implemented in the
SARscape modules (software version 5.2.1) available under ENVI plat­
form (software version 5.3). For the Miami Beach study area, 95 inter­
ferometric pairs were connected by setting the maximum perpendicular
and temporal baselines respectively to 55% (relative to the critical
baseline value, Table S3) and 1,000 days (Fig. 2A). For the Norfolk study
area, 182 interferometric pairs were connected by setting the baselines
to 60% (relative to the critical baseline value, Table S4) and 1,200 days
(Fig. 2B). A multi-look factor of 1 in slant range and 5 in azimuth di­
rections was applied to obtain a ground pixel resolution of ~20 m by 20
m. The interferometric phase related to the topography was removed
using the 1 arc-second Shuttle Radar Topography Mission (SRTM) DEM.
Orbital inaccuracies are compensated by using ESA’s Precise Orbits
(PRC) data. Spectral shift and common Doppler bandwidth filtering are
applied during the interferogram generation to minimize decorrelation
effects (Gatelli et al., 1994). The noise and signal frequencies are
smoothed in a selective way by using the Goldstein adaptive filter
(Goldstein and Werner, 1998). The filter strength is based on the local
coherence and is controlled by the “α” parameter that was set to 0.3–1.0
(minimum values) and 2.5–3.0 (maximum values) respectively for
Miami Beach and Norfolk. The interferograms are co-registered to the
geometry of a master image and unwrapped using the Delaunay Mini­
mum Cost Flow (MCF) algorithm (Costantini, 1998). An unwrapping
coherence threshold of 0.30 was set for both data sets. Residual orbital
Fig. 2. SBAS connection networks. Interferometric pairs generated for (A) Miami Beach from 1993 to 1999 and (B) Norfolk (VA) from 1992 to 1998. The red dots are
the images used as master. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
3
Ocean and Coastal Management 187 (2020) 105078
inaccuracies, phase offsets and/or phase ramps are removed by using a
cubic polynomial refinement on selected ground control points (Figs. S2
and S3). The ground control points used in this study are not calibrated
with ground velocities obtained with other geodetic methods such as
GPS, and hence the produced maps of relative subsidence do not take
into account any regional-scale deformation. The residual height
correction factors and the average displacement rates are retrieved by
inverting a linear system via the Singular Value Decomposition (SVD)
method (Strang, 1988). The Atmospheric Phase Screen (APS) is then
calculated and removed by applying spatial (low-pass) and temporal
(high-pass) filters. The corrections are carried out assuming that the
atmospheric phase delays are correlated in space and uncorrelated in
time (Emardson et al., 2003). For both data sets, the low-pass and
high-pass filters were set respectively to 1200 m and 365 days. The final
products consisting of mean velocities and displacements time series are
calculated and geocoded to the WGS 1984 UTM Zone 17N (Miami
Beach) and 18N (Norfolk) projected coordinate system. During the SBAS
processing chain the multi-temporal coherence is estimated for each
pixel by evaluating the interferometric correlation in the entire in­
terferograms stack. The coherence jγj between two coregistered complex
SAR images y1 and y2 is calculated as the maximum likelihood estimator
of the coherence magnitude over an estimation window of N pixels
(Seymour and Cumming, 1994)
�PN ðnÞ *ðnÞ �
� y y �
ffiffiffiffiffiffiffiffi1ffiffiffiffiffiffiffi2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0 � jγj � 1
jγj ¼ qffiffiffiffiffiffiffiffiffiffiffiffin¼1
PN ðnÞ 2 PN ðnÞ 2
n¼1 jy1 j n¼1 jy2 j
The multi-temporal coherence (denoted as γ for simplicity) can be
used as a measure of the accuracy of the interferometric phase (Hanssen,
2001). The relative uncertainty (σ) (Sarmap, 2014) of the mean
InSAR-derived velocities is then internally calculated for each pixel as
(simplified from: Bamler and Just, 1993)
sffiffiffiffiffiffiffiffiffiffiffiffiffi
λ 1 γ2
σ¼
4π 2γ2
where λ is the signal wavelength. The obtained uncertainty values are
between 0.2 mm/yr and 1.5 mm/yr (mean value of 0.6 mm/yr) for
Miami Beach (Fig. S4), and between 0.1 mm/yr and 2.4 mm/yr (mean
value of 0.6 mm/yr) for Norfolk (Fig. S5). Assuming that the ground
movements in both study areas do not present strong horizontal com­
ponents, the mean LOS velocities are projected to the vertical direction
taking into account the incidence angle of each measured point. The
final results were classified in a Geographic Information System (GIS)
(ArcGIS 10.6 software) using a colour scale from blue to red, in which
positive values, in blue, correspond to a decrease of the sensor to target
slant range distance (i.e. relative uplift), while negative values, in red,
correspond to an increase of the sensor to target slant range distance (i.e.
relative subsidence). The points considered as stable, in green, are the
ones with velocities of �1 mm/yr, close to the accuracy obtainable with
the ERS-1/2 satellites (Crosetto et al., 2008).
S. Fiaschi and S. Wdowinski
4. Results
Miami Beach is a densely populated barrier island, located between
Biscayne Bay and the Atlantic Ocean (Fig. 1b). It is roughly 10 km long
and 2.5 km wide at its widest point. InSAR time series results reveal a
patch-like pattern of coherent velocity observations, reflecting contri­
butions from variable scattering environments and high variability of
tropospheric delay. Coherent results occur over the built environment,
whereas incoherent results occur over open areas and some of the built
environment. The point density obtained with the SBAS approach is of
924 points/km2. Among the 18,949 measured points, 1555 present ve­
locities below 1.0 mm/yr, while only 11 of them have velocities below
3 mm/yr (Table S5). The results indicate that most of the city (~97%)
was stable during the 1993–1999 observation period (green in Fig. 3A).
However, several localized subsiding areas were detected mostly in the
western part of the city (yellow in Fig. 3A). These areas typically consist
of single-family houses that were built on reclaimed wetlands (e.g.
southernmost black circles in Fig. 3A). In some locations, as in the
eastern part of the city, the detected subsidence is of a 12-story high
condominium building (northernmost black circle in Fig. 3A). The
detected subsidence rate is in the 1–3 mm/yr range, with uncertainty
level of 0.6–0.8 mm/yr. Although higher subsidence velocities up to 3.8
mm/yr are registered in the artificial islands located west of the city,
phase unwrapping errors leading to higher uncertainties cannot be
excluded in such areas.
The city of Norfolk, Virginia, is located in lower Chesapeake Bay,
which is the largest estuary in the U.S (Fig. 1A). The city is surrounded
by water from the Bay in the north and secondary rivers and estuaries in
the west and south (Fig. 1C). Since the entire Chesapeake Bay area is
subject to regional subsidence, the results obtained from the SBAS InSAR
processing over Norfolk are the expression of the differential movements
of the measured points relative to the subsiding land surface. To obtain
the total subsidence in the area, the contribution of the regional subsi­
dence obtained from GPS observations from 2009 to 2017 (1–2.4 mm/
4
Ocean and Coastal Management 187 (2020) 105078
yr) (Bekaert et al., 2017) must be added to the InSAR measurements. The
InSAR analysis over Norfolk extends well beyond the city limits (Fig. 4)
and covers also the nearby GPS sites. However, the results here reported
refer, unless otherwise specified, only to the area of the municipality
(Fig. 1C). Thanks to the SBAS technique, we obtained very high point
density (2140 points/km2) and covered almost the entire territory of
Norfolk, except for the areas characterized by very low coherence cor­
responding to water bodies and dense vegetation (Fig. 4). The results,
show that most of the city lies over a relatively stable territory while
only few localized areas are affected by subsidence up to 6.0 mm/yr.
Among the 313,052 measured points, 74,831 registered subsidence
greater than 1.0 mm/yr (24%), and only 1106 greater than 3 mm/yr
(0.3%). Four main locations are here reported as examples of localized
subsidence in the Norfolk area (Figs. 4 and 5): A) Willoughby Spit and
tunnel island; B) Craney Island and Lamberts Point terminal; C) East
Beach residential area; D) Hampton Road Beltway. The extent of the
subsiding areas is generally limited to localized residential or industrial
sites; only in the case of Norfolk International Airport and its sur­
roundings the subsidence extent is larger (Fig. 5D). Subsidence in this
area also intersects the Hampton Road Beltway for a total length of
around 6000 m. In the four selected locations, most of the subsiding
points registered velocities between 1.0 mm/yr and 3.0 mm/yr,
while only a smaller percentage had higher subsidence rates, up to 6.0
mm/yr (Table S6). The measurement uncertainties are in the range of
0.1–2.4 mm/yr.
A section of the Hampton Roads Beltway close to the Norfolk Inter­
national Airport shows subsidence rates ranging from 1.0 mm/yr to 4.0
mm/yr (Fig. 5D). Outside the study area, around 9 km west of the
Norfolk city limits, another 4000 m long section of the Hampton Road
Beltway is affected by higher subsidence rates up to 8.3 � 1.0 mm/yr
(Fig. 4). Localized subsidence occurred also in the East Beach residential
area (Fig. 5C), north of the city. Here, the subsidence rates can reach 3.3
mm/yr. Similar rates of subsidence affected Willoughby Spit in the
northern section of the city, whereas the west flank of the man-made
Fig. 3. Vertical velocity map over Miami Beach from
1993 to 1999 and displacement time series. (A)
vertical velocity map obtained with the SBAS tech­
nique. The black circles mark the location of the
extracted displacement time series. Red lines mark
the location of roads affected by flooding during
1998–2012 (data from Wdowinski et al., 2016). (B)
Vertical displacement time series of the four selected
points. Base image source: Esri, DigitalGlobe, Geo­
Eye, Earth Geographics, CNES/Airbus DS, USDA,
USGS, AeroGRID, IGN, and the GIS User Community.
(For interpretation of the references to colour in this
figure legend, the reader is referred to the Web
version of this article.)
S. Fiaschi and S. Wdowinski
island at the entrance to the Hampton Roads Bridge–Tunnel (HRBT)
registered subsidence up to 6.3 mm/yr (Fig. 5A). Subsidence up to 4.1
mm/yr affected also parts of the Lamberts Point coal terminal, whereas
the subsidence is more evident in the near Craney Island where the
entire Oily Waste Treatment Plant subsided with rates reaching 5.1 mm/
yr (Fig. 5B). The analysis of the displacement time series extracted in the
four main areas (Fig. 5), shows a similar deformation trend that appear
to be linear in each of the selected points. The non-uniform temporal
distribution of the images, with a gap of information between 09/1993
and 04/1995, and the production of decorrelation effects can explain the
occurrence of anomalous displacement values that are visible, for
example in P4.
5. Discussion
InSAR time series analysis of the two coastal communities, Miami
Beach and Norfolk, reveals that during the observation period of
1992–1999, both communities were subjected to localized differential
subsidence at rates of 1–3 mm/yr in Miami Beach and 1–6 mm/yr in
Norfolk. In Miami Beach, subsidence is limited to several small areas
(~3% of the analysed area; Fig. 3), mostly located in the western side of
the city, which was built on reclaimed wetlands. The rest of the city
remained stable, which agree with the two years of continuous GPS
observations in Miami Beach (Station: FLMB) and four other station
locations in southeast Florida. The geology of the area is, in fact,
constituted by a plateau of karst limestone (Miami Limestone Forma­
tion) sitting over the Florida Platform (Lane, 1994). Although no
regional subsidence has been recorded in south Florida, localized sub­
sidence can occur as consequence of karst activity and compaction of
unconsolidated sediments. In Norfolk, differential subsidence occurred
over wider areas (~20% of the municipality), as wide as 7–8 km, but
also in smaller areas with dimension of tens to hundreds of meters
(Figs. 4 and 5). Subsidence affected also critical infrastructure such as
the Hampton Road Beltway (Fig. 5D). This road is one of the main
designated hurricane evacuation routes for people in vehicle to flee the
Virginia Beach (East of Norfolk) and Norfolk areas, as reported from the
Virginia Department of Emergency Management. The presence of sub­
siding areas along the route may increase the flooding hazard in case of
storm surges and can be a significant aspect to be considered by the
5
Ocean and Coastal Management 187 (2020) 105078
Fig. 4. Vertical velocity map over Norfolk (VA) from
1992 to 1998. The white squares are the locations of
the four areas affected by land subsidence reported
in Fig. 5. Purple triangles are the locations of the GPS
stations available in the area. The light blue triangle
is the location of the Sewells Point tide gauge station.
Base image source: Esri, DigitalGlobe, GeoEye, Earth
Geographics, CNES/Airbus DS, USDA, USGS, Aero­
GRID, IGN, and the GIS User Community. (For
interpretation of the references to colour in this
figure legend, the reader is referred to the Web
version of this article.)
Authorities for planning evacuation strategies. Our Miami Beach results
are first to report land subsidence in the city. However, land subsidence
in Norfolk was reported previously, both in term of regional component
(Karegar et al. 2016, 2017) and localized differential component
(Bekaert et al., 2017). Although most of Norfolk seems stable (green in
Fig. 4), the entire city subsides at a rate of 1.5–2.5 mm/yr as part of the
GIA process (Bekaert et al., 2017; Karegar et al. 2016, 2017). Thus, we
added the regional component derived from the LOY2 GPS during
02/2009–04/2017 ( 1.7 � 0.9 mm/yr) to our observed relative subsi­
dence (1–6 mm/yr) and presented it in Fig. S6.
The previous study of Bekaert et al. (2017) used GPS observations in
the lower Chesapeake Bay area to calibrate their InSAR results, which
were derived from ALOS-1 observations acquired during 2007–2011.
Their results revealed a very diffuse subsidence pattern occurring
throughout most of the city with rates ranging 0–10 mm/yr. However,
the limited number of SAR data and the 4-year observation period used
in their study yielded very high uncertainty rates (5–8 mm/yr), which
were not sufficient for generating accurate subsidence maps (Bekaert
et al., 2017). Our results, however, rely on more observations and a
longer acquisition period (6–7 years) and, hence, yielded a more accu­
rate subsidence map (uncertainties up to 2.4 mm/yr) showing localized
subsidence patterns.
Localized subsidence typically occurs in response to shallow sub­
surface processes as sediment compaction, often in response to
groundwater extraction (Amelung et al., 1999; Galloway and Burbey,
2011), and soil settlement in reclaimed areas (e.g., Kim et al., 2008). The
rate of localized subsidence can vary over time, depending on the con­
trolling mechanisms. Subsidence rate due to groundwater exploitation is
affected by extraction and recharge histories, as well as by the defor­
mation characteristics of the aquifer system. Subsidence rate due to soil
settlement in reclaimed areas also varies in time, as the rate of soil
compaction decreases over time. Despite the time dependent nature of
localized subsidence, it can be characterized by linear rates, as presented
in this study, reflecting a mean, uniform rate of subsidence that occurred
during the observation period. The application of the observed rates
beyond the observation periods is a good first assumption, as the
deformation processes tend to vary at slow rates, especially if no other
observations are available. However, the extension of the observed
subsidence rates beyond the observation period should be done with
S. Fiaschi and S. Wdowinski
caution.
The high spatial resolution of InSAR time series analysis can be used
to evaluate stability of buildings, as well as point measurement of
geodetic instruments, such as continuous GPS stations or tide gauges, as
previously done in the Venice Lagoon (Italy) (Bock et al., 2012) and in
Berst (France) (Poitevin et al., 2019). Here, we use the InSAR-derived
relative velocity map of Norfolk to evaluate the stability of the Sewells
Point tide gauge, which is used to estimate the rate of relative SLR for
Norfolk, and the LOY2 continuous GPS station (Fig. 4), which is used to
estimate the regional land subsidence component (1.7 � 0.9 mm/yr).
Unfortunately, we were unable to evaluate the stability of the LOYZ and
DRV5/6 GPS stations (which measure regional land subsidence respec­
tively of 1.8 � 0.8 mm/yr and 1.9 � 0.5 mm/yr), because both stations
are surrounded by vegetated areas, which do not maintain sufficient
interferometric coherence in the C-band ERS-1/2 data to obtain reliable
results. We are also unable to directly determine the stability of the
Sewells Point tide gauge, as the gauge is located at the end of a peer,
6
Ocean and Coastal Management 187 (2020) 105078
Fig. 5. Main areas affected by subsidence in Norfolk
(VA). (A) Willoughby Spit and tunnel island. (B)
Craney Island (left) and Lamberts Point terminal
(right). (C) East Beach residential area. (D) Hampton
Road Beltway. The white circles mark the location of
the extracted displacement time series. (P1–P4)
Vertical displacement time series extracted from four
selected points in Norfolk (VA). Base image source:
Esri, DigitalGlobe, GeoEye, Earth Geographics,
CNES/Airbus DS, USDA, USGS, AeroGRID, IGN, and
the GIS User Community.
which does not provide a good scattering environment and shows patchy
areas of slight uplift (insert in Fig. 4). However, the scatterers located
within the broader area of the Naval Base indicate relative stability of
the area (average movement rate of 28,401 points is 0.5 � 0.6 mm/yr).
After correcting the relative InSAR subsidence field by the regional
subsidence component of ~1.7 mm/yr (Fig. S6), the Naval Base exhibits
an average subsidence rate of ~1.2 mm/yr, which explains the high
relative SLR at Sewells Point (4.6 � 0.3 mm/yr) compared to the global
mean SLR (3.2 � 0.3 mm/yr). Our results also indicate that the LOY2
GPS station is stable based on statistical analysis of 18 InSAR points
found within 50 m buffer distance from the GPS station revealing rela­
tive average motion of 0.4 � 0.4 mm/yr.
We did not conduct a similar stability analysis for Miami Beach,
because no tide gauge or GPS stations were operating in the study area
during the ERS-1/2 measurement period. The Miami Beach tide gauge
operated during the years 1931–1981 and then was relocated to Virginia
Key in 1994. The one continuous GPS station operating in Miami Beach
S. Fiaschi and S. Wdowinski
(FLMB) started operation in 2017 and, insofar, its record indicates no
measurable subsidence.
Coastal flooding hazard assessments often rely on the rate of relative
SLR for forecasting inundation level and/or flooding frequencies in
coastal communities. For example, Burgos et al. (2018) calculated future
nuisance flooding frequency in Norfolk using rate of relative SLR of 4.6
� 0.3 mm/yr derived from the 1927–2016 Sewells Point’s tide gauge
record. These results are representative for most of the city, but not for
areas subjected to localized subsidence. In those areas the rate of relative
SLR is higher and can reach level of 8–9 mm/yr, according to the ob­
tained results. In particular, localized subsidence may naturally occur in
low elevation areas that are more vulnerable to flooding. Based on a
LiDAR DEM acquired in 2013, we calculated that 614 of the total sub­
siding points in the Norfolk municipality lie at an elevation <1 m above
North American Vertical Datum of 1988 (NAVD88), which is < 1.08 m
above Mean Sea Level. These 614 points represent an area of ~25 ha
(0.17% of the territory) in which the flooding hazard can be higher due
to the combination of local subsidence and low elevation. In Miami
Beach, some of the localized subsiding areas are located hundreds of
meters inland of the shoreline (Fig. 3), suggesting a negligible contri­
bution for increasing the rate of relative SLR at these sites. However,
these localized subsiding areas can increase the likelihood of
rain-induced flooding, which occur during heavy rain events. The
comparison between the InSAR results and the flooding events reported
between 1998 and 2015 (Fig. 6) shows that 59% and 30% of the
detected subsiding points in Miami Beach and Norfolk municipalities,
respectively, are within 200 m of areas already subjected to flooding.
This suggests that a contribution of the detected subsidence to the more
recent flooding events in these areas is possible, but further and more
recent data are required to confirm such statement.
In the current study, we used ERS-1/2 data acquired during a 7-year
long period, which allow us to obtain results with uncertainty lower
than 1 mm/yr in most areas (Figs. S4 and S5). However, in some loca­
tions, uncertainty reached 2–3 mm/yr level, suggesting caution in the
7
Ocean and Coastal Management 187 (2020) 105078
evaluation of the velocities in these areas. Another issue that can limit
the significance of our results is nature of the measured objects on the
surface. SAR measurements in urban areas are dominated by double-
bounce backscattering from buildings (Hong and Wdowinski, 2014;
Yamaguchi et al., 2005) and, hence, represent mainly building move­
ments. Thus, subsiding areas with stable buildings, which are anchored
at depth by piling to a stable stratum, will reveal in InSAR analysis as
stable, non-subsiding surface conditions. As an example, such differen­
tial subsidence behaviour was detected in Mexico City indicating
apparent uplift of large, well anchored building or structures, as Metro
stations, with respect to their surroundings (Osmanoglu et al., 2011;
Solano-Rojas et al., 2019). Therefore, our reported subsidence results
may represent movement of buildings and not of the actual surface.
Here we focused on Miami Beach and Norfolk because both have
been subjected in the past couple of decades to increased frequency of
nuisance flooding, which caused property damage and affected daily
life. However, the importance of our study extends well beyond these
two coastal areas, because similar deformation processes affect many
other coastal communities. With the projected increase of sea level,
additional communities along the US Atlantic coast and in other coastal
areas worldwide will be subjected to more frequent nuisance flooding.
Thus, improved management of coastal flooding hazard in most coastal
communities should take into account studies and assessments of ver­
tical land movements, which include contributions from both regional
and local land subsidence. The regional component is best evaluated
using GPS measurements, whereas the local subsidence is best measured
with InSAR in urban areas and GPS in open areas. Thus, future assess­
ments of coastal flooding hazards should include measurements and
analyses of both GPS and InSAR data. Recently, the state of Florida
initiated a coastal subsidence monitoring project in Miami-Dade County,
which implements the recommended dual GPS-InSAR monitoring
approach.
Our InSAR time series analysis has some limitations in terms of
observation period and measurement uncertainty. Our results are based
Fig. 6. Comparison between areas affected by sub­
sidence detected with InSAR (vertical velocity < 1
mm/yr, red) and the flooding reported (blue) in
Miami Beach (data from Wdowinski et al., 2016) (A)
and Norfolk (data from http://va-norfolk.civicplus.
com) (B) municipalities. Data in (B) are super­
imposed to the LiDAR elevation map (NAVD88).
Base image source: Esri, DigitalGlobe, GeoEye, Earth
Geographics, CNES/Airbus DS, USDA, USGS, Aero­
GRID, IGN, and the GIS User Community. (For
interpretation of the references to colour in this
figure legend, the reader is referred to the Web
version of this article.)
S. Fiaschi and S. Wdowinski
on ERS-1/2 data acquired more than two decades ago and, hence, reflect
localized subsidence occurred during the observation time period
(1992–1999). As explained above, the application of the observed rate
beyond the observation period should be done with caution. Unfortu­
nately, follow up SAR missions, as ENVISAT and ALOS-1, did not acquire
a sufficient number of images over the US Atlantic coast to obtain reli­
able velocity observations needed for assessing coastal subsidence at the
1–2 mm/yr uncertainty level. The currently operating Sentinel-1
mission acquires SAR data over the US Atlantic coast in a systematic
manner since 2016. The currently available Sentinel-1 data expand over
a period of three years, which is still not sufficient to detect surface
movements at the desired 1–2 mm/yr uncertainty level (Havazli and
Wdowinski, 2017). However, this uncertainty level will be reached as
soon as Sentinel-1 acquisitions will cover a period greater than four/five
years (Havazli and Wdowinski, 2017). Considering the great potential of
Sentinel-1 data to detect ground movements over large scales at
improved resolution in comparison with past-generation satellites,
future work will consist in upscaling this study to a larger number of
communities along the US Atlantic coastline, and assessing the contri­
bution of the more recent land subsidence to flooding hazard in such
areas.
6. Concluding remarks
In this study, SAR images acquired by the ERS-1/2 satellites between
1992 and 1999 are used to measure local land subsidence in two com­
munities located along the US Atlantic coastline, Miami Beach (FL) and
Norfolk (VA). Both communities have been affected by increasing fre­
quency of flooding events, in Miami Beach mainly due to increasing rate
of sea level rise, while in Norfolk mainly due to regional subsidence as
consequence of GIA and groundwater exploitation. Our InSAR results
show that in both communities land subsidence contributes locally to
higher relative sea level rise rates. In Miami Beach, land subsidence
occurred only locally and at very low rates, between 1 mm/yr and 3
mm/yr. In Norfolk, land subsidence affected a larger portion of the
territory with rates up to 6 mm/yr (8 mm/yr if considering both relative
and regional components). This study shows the importance of consid­
ering land subsidence as one of the principal factors that affect the
spatial distribution of flooding events and their frequency in low-lying
coastal communities. The information provided by the InSAR analysis
are valuable to local authorities for improving flood hazard assessments
and for helping the development and implementation of more efficient
flood emergency response and recovery plans.
Although this study provides an unprecedented overview of the land
motion occurring in the Miami and Norfolk communities, our InSAR
results are limited to a period of time of more than 20 years ago. For this
reason, future work will focus on the application of multi-temporal
InSAR techniques to more recent Sentinel-1 data for the monitoring of
a larger extent of the communities along the US Atlantic coast.
Data Availability
ERS-1/2 data are provided by ESA through the EOLi-SA (Earth
Observation Link) client for Earth Observation Catalogue Service. The
SRTM DEM is provided by the United States Geological Survey (USGS)
though the EarthExplorer online catalogue (https://earthexplorer.usgs.
gov/). LiDAR data of Norfolk (VA) are provided by NOAA Office for
Coastal Management (https://coast.noaa.gov).
Funding
This research was partly conducted during S.F. PhD at the Depart­
ment of Geosciences of the University of Padua, Italy. S.W. thanks the
Florida Office of Insurance Regulations for supporting the Miami Beach
component of the research. This is contribution number 938 from the
Southeast Environmental Research Center in the Institute of
8
Ocean and Coastal Management 187 (2020) 105078
Environment at Florida International University.
Author contributions
S.F. carried out the InSAR analysis. S.W. helped with the interpre­
tation of the results. Both authors contributed in the writing of the
manuscript.
Data and materials availability
All data needed to evaluate the results and conclusions reported in
the paper are present in the main text or in the Supplementary Materials.
If necessary, further data may be requested from the authors.
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.ocecoaman.2019.105078.
References
Aerts, J.C.J.H., Botzen, W.J.W., Emanuel, K., Lin, N., de Moel, H., Michel-Kerjan, E.O.,
2014. Evaluating flood resilience strategies for coastal megacities. Science 344,
472–474.
Amelung, F., Galloway, D.L., Bell, J.W., Zebker, H.A., Laczniak, R.J., 1999. Sensing the
ups and downs of Las Vegas: InSAR reveals structural control of land subsidence and
aquifer-system deformation. Geology 27, 483–486.
Bamler, R., Just, D., 1993. Phase statistics and decorrelation in SAR interferograms. In:
Geoscience and Remote Sensing Symposium. IGARSS, pp. 980–984.
Bekaert, D.P.S., Hamlington, B.D., Buzzanga, B., Jones, C.E., 2017. Spaceborne synthetic
aperture radar Survey of subsidence in Hampton roads, Virginia (USA). Sci. Rep. 7.
Berardino, P., Fornaro, G., Lanari, R., Sansosti, E., 2002. A new algorithm for surface
deformation monitoring based on small baseline differential SAR interferograms.
IEEE Trans. Geosci. Remote Sens. 40, 2375–2383.
Bock, Y., Wdowinski, S., Ferretti, A., Novali, F., Fumagalli, A., 2012. Recent subsidence
of the Venice Lagoon from continuous GPS and interferometric synthetic aperture
radar. Geochem. Geophys. Geosyst. 13.
Burgos, A.G., Hamlington, B.D., Thompson, P.R., Ray, R.D., 2018. Future nuisance
flooding in Norfolk, VA, from astronomical tides and annual to decadal internal
climate variability. Geophys. Res. Lett. 45, 12432–12439.
Chaussard, E., Amelung, F., Abidin, H., Hong, S.H., 2013. Sinking cities in Indonesia:
ALOS PALSAR detects rapid subsidence due to groundwater and gas extraction.
Remote Sens. Environ. 128, 150–161.
Costantini, M., 1998. A novel phase unwrapping method based on network
programming. IEEE Trans. Geosci. Remote Sens. 36, 813–821.
Crosetto, M., Monserrat, O., Adam, N., Parizzi, A., Bremmer, C., Dortland, S., Hanssen, R.
F., Van Leijen, F.J., 2008. Validation of existing processing chains in TerraFirma
stage 2-Final Report. In: GMES TERRAFIRMA ESRIN, p. 15.
Dixon, T.H., Amelung, F., Ferretti, A., Novali, F., Rocca, F., Dokka, R., Sella, G., Kim, S.
W., Wdowinski, S., Whitman, D., 2006. Subsidence and flooding in new Orleans.
Nature 441, 587–588.
Eggleston, J., Pope, J., 2013. Land subsidence and relative sea-level rise in the southern
Chesapeake Bay region. In: U.S. Geological Survey Circular.
Emardson, T.R., Simons, M., Webb, F.H., 2003. Neutral atmospheric delay in
interferometric synthetic aperture radar applications: statistical description and
mitigation. Journal of Geophysical Research-Solid Earth 108.
Engelhart, S.E., Horton, B.P., 2012. Holocene sea level database for the Atlantic coast of
the United States. Quat. Sci. Rev. 54, 12–25.
Ezer, T., 2013. Sea level rise, spatially uneven and temporally unsteady: why the US East
Coast, the global tide gauge record, and the global altimeter data show different
trends. Geophys. Res. Lett. 40, 5439–5444.
Galloway, D.L., Burbey, T.J., 2011. Review: regional land subsidence accompanying
groundwater extraction. Hydrogeol. J. 19, 1459–1486.
Gatelli, F., Guarnieri, A.M., Parizzi, F., Pasquali, P., Prati, C., Rocca, F., 1994. The wave-
number shift in sar interferometry. IEEE Trans. Geosci. Remote Sens. 32, 855–865.
Goldstein, R.M., Werner, C.L., 1998. Radar interferogram filtering for geophysical
applications. Geophys. Res. Lett. 25, 4035–4038.
Hanssen, R.F., 2001. Radar Interferometry: Data Interpretation and Error Analysis.
Springer, Dordrecht.
Havazli, E., Wdowinski, S., 2017. The effect of temporal resolution due to atmospheric
phase delay on minimum detectable signal in InSAR time series: application to slow
deformation over socorro magma body. In: FRINGE. Finland, Helsinki.
S. Fiaschi and S. Wdowinski
Hong, S.H., Wdowinski, S., 2014. Double-bounce component in cross-polarimetric SAR
from a new scattering target decomposition. IEEE Trans. Geosci. Remote Sens. 52,
3039–3051.
Karegar, M.A., Dixon, T.H., Engelhart, S.E., 2016. Subsidence along the atlantic coast of
north America: insights from GPS and late Holocene relative sea level data. Geophys.
Res. Lett. 43, 3126–3133.
Karegar, M.A., Dixon, T.H., Malservisi, R., Kusche, J., Engelhart, S.E., 2017. Nuisance
flooding and relative sea-level rise: the importance of present-day land motion. Sci.
Rep. 7.
Kemp, A.C., Bernhardt, C.E., Horton, B.P., Kopp, R.E., Vane, C.H., Peltier, W.R.,
Hawkes, A.D., Donnelly, J.P., Parnell, A.C., Cahill, N., 2014. Late Holocene sea- and
land-level change on the US southeastern Atlantic coast. Mar. Geol. 357, 90–100.
Kim, S.W., Wdowinski, S., Dixon, T.H., Amelung, F., Won, J.S., Kim, J.W., 2008. InSAR-
based mapping of surface subsidence in Mokpo City, Korea, using JERS-1 and
ENVISAT SAR data. Earth Planets Space 60, 453–461.
Kirshen, P., Knee, K., Ruth, M., 2008. Climate change and coastal flooding in Metro
Boston: impacts and adaptation strategies. Clim. Change 90, 453–473.
Kleinosky, L.R., Yarnal, B., Fisher, A., 2007. Vulnerability of Hampton Roads, Virginia to
storm-surge flooding and sea-level rise. Nat. Hazards 40, 43–70.
Lane, E. (Ed.), 1994. Florida’s Geological History and Geological Resources Special
Publication No. 35. Department of Environmental Protection, Florida Geological
Survey, Tallahassee, FL.
Nikitina, D., Kemp, A.C., Engelhart, S.E., Horton, B.P., Hill, D.F., Kopp, R.E., 2015. Sea-
level change and subsidence in the Delaware Estuary during the last-2200 years.
Estuar. Coast Shelf Sci. 164, 506–519.
Ocean and Coastal Management 187 (2020) 105078
Osmanoglu, B., Dixon, T.H., Wdowinski, S., Cabral-Cano, E., Jiang, Y., 2011. Mexico City
subsidence observed with persistent scatterer InSAR. Int. J. Appl. Earth Obs. Geoinf.
13, 1–12.
Poitevin, C., Woppelmann, G., Raucoules, D., Le Cozannet, G., Marcos, M., Testut, L.,
2019. Vertical land motion and relative sea level changes along the coastline of Brest
(France) from combined space-borne geodetic methods. Remote Sens. Environ. 222,
275–285.
Sarmap, 2014. SARscape 5.1 help. Available online at: https://www.harrisgeospatial.
com/docs/pdf/sarscape_5.1_help.pdf.
Sella, G.F., Stein, S., Dixon, T.H., Craymer, M., James, T.S., Mazzotti, S., Dokka, R.K.,
2007. Observation of glacial isostatic adjustment in "stable" North America with
GPS. Geophys. Res. Lett. 34.
Seymour, M.S., Cumming, I.G., 1994. Maximum likelihood estimation for SAR
interferometry. In: Proceedings of IGARSS ’94 - 1994 IEEE International Geoscience
and Remote Sensing Symposium, pp. 2272–2275 (Pasadena, CA, USA).
Solano-Rojas, D., Wdowinski, S., Cabral-Cano, E., Osmano� glu, B., 2019. Geohazard
assessment of Mexico City’s Metro System derived from SAR interferometry
observations. Science Advances. Submitted for publication.
Strang, G., 1988. Linear Algebra and its Appications. Orlando, FL, USA.
Wdowinski, S., Bray, R., Kirtman, B.P., Wu, Z.H., 2016. Increasing flooding hazard in
coastal communities due to rising sea level: case study of Miami Beach, Florida.
Ocean Coast Manag. 126, 1–8.
Yamaguchi, Y., Moriyama, T., Ishido, M., Yamada, H., 2005. Four-component scattering
model for polarimetric SAR image decomposition. IEEE Trans. Geosci. Remote Sens.
43, 1699–1706.
Zervas, C.E., 2009. Sea Level Variations of the United States. NOAA Technical Report
NOS CO-OPS 053, pp. 1854–2006.